Social Support for Divorced Fathers' Parenting: Testing a Stress-Buffering Model

Abstract

A stress-buffering hypothesis for parenting was tested in a county-representative sample of 218 divorced fathers. Social support for parenting (emergency and nonemergency child care, practical support, financial support) was hypothesized to moderate effects of stress (role overload, coparental conflict, and daily hassles) on fathers’ quality parenting. No custody fathers relied more on relatives compared with custodial fathers, who relied more on new partners for parenting support. No differences by custody status were found on levels of support or parenting over time. Parenting support buffered effects of change in role overload and coparenting conflict on coercive parenting and buffered effects of change in daily hassles on prosocial parenting. Buffer effects were more predictive over time. Implications for practice and preventive intervention strategies are discussed.

Ecological models on the continuum of child maltreatment to effective quality parenting have identified interpersonal support as a buffer to life stresses and as a key determinant of effective parenting (Belsky, 1984; Simons & Johnson, 1996). Conversely, social isolation is a risk factor (Belsky & Vondra, 1989; Hutchings, Midence, & Nash, 1997). Social support is also a key factor promoting father involvement for residential (Lamb, 2002) and non-residential fathers (Braver et al., 1993; Doherty, Kouneski, & Erickson, 1998). Yet, despite its theoretical and practical significance, social support specifically for father’s parenting has rarely been studied and studied less so in the context of divorce. The majority of empirical studies have primarily focused on direct and indirect benefits for divorced mothers’ parenting (Bretherton, Walsh, & Lependorf, 1996; DeGarmo & Forgatch, 1997; Simons, Beaman, Conger, & Chao, 1993). Therefore, little is known about how divorced fathers’ use and benefit from social support for parenting, even with the fairly substantial literature recently emerging on father involvement. Further, research on parenting supports for divorced fathers has focused mainly on occupational and policy supports for father involvement, typically measured as contact with children (DeMaris & Greif, 1997; Greif, 1995).

Therefore, understanding effects of social support is particularly important for divorced fathers because mothers, fathers, and children all benefit from quality postdivorce father involvement (Hetherington, Bridges, & Insabella, 1998; King & Sobolewski, 2006) and more father involvement is associated with better quality visitation (Arditti & Keith, 1995). Support is also salient because separation results in greater psychological distress in fathers than in mothers and is more pronounced in the initial stages of divorce (Bloom, Asher, & White, 1978; Jacobs, 1982).

Focusing on parenting support, we attempted to advance prior research on father involvement in several ways. First, we focused on divorced fathers’ coercive and prosocial parenting behaviors, moving beyond simple measures of fathers’ contact with children. It is the quality of divorced fathers’ parenting, more so than the mere quantity of visits, that is predictive of family adjustment (Amato, 1993; Hetherington et al., 1998; Lamb, 2002).

Second, beyond general social support, we addressed to whom fathers turn for specific parenting needs. Because parenting needs differ as a function of custodial responsibilities, we examined sources of support by custody status. Research shows that shared custody fathers are more satisfied with parenting arrangements compared with noncustodial fathers (Arditti, 1992), and shared custody parents, in general, report less stress and conflict than sole custody parents (Bauserman, 2002). Because we were interested in parenting behaviors that may have a developmental impact, we defined shared custody as joint legal or joint physical custody in which the father has contact and parenting interaction with the child.

Third, we employed a theoretically grounded stress-buffering model to test whether social support buffers the potentially negative impact of stressors on divorced fathers’ parenting over time. Three common stressors were tested: conflict with the former spouse, family stressors, and fathers’ role overload. The theoretical model is shown in Figure 1 with hypotheses briefly developed below. The general model suggests that common stressors will have a negative impact on parenting quality for divorced fathers; however, this negative impact will be lessened for fathers with higher levels of social support.

Figure 1.

Theoretical Model of Parenting Support as a Stress Buffer for Divorced Fathers’ Parenting.

Family Stressors and Fathering After Divorce

Divorce is associated with a high degree of stress. We briefly outline three common stressors for divorced fathers hypothesized to be associated with parenting practices: conflict with the former spouse, role overload, and daily life and family stressors. We note that constructs measured in the present study are not an exhaustive or comprehensive list of potential stressors.

Coparenting conflict

The experience of coparenting conflict is particularly stressful for both divorced parents and is associated with higher levels of psychological distress for all fathers and lower levels of involvement for joint and no custody fathers (Braver et al., 1993; De Luccie, 1995). Braver, Griffin, and Cookston (2005) reported that most divorcing parents experience initially high levels of conflict for up to 3 years, after which couples tend to disengage from protracted conflict and either engage in parallel parenting or cooperative coparenting with roughly one quarter maintaining conflict. Many couples also establish a cooperative coparenting relationship to provide a united front in the best interest of their child (Ahrons & Miller, 1993). Therefore, we hypothesized that conflict would be associated with lower levels of quality parenting.

Role overload

Parents must establish new rules for parenting together in new family structures, while at the same time relinquishing their roles as marital partners (Emery, 1994). The majority of mothers must also adjust to the role of sole custodian, taking on primary responsibility for household management and parenting needs. Although fathers are increasingly more involved than in the past (Day & Lamb, 2004), historically speaking, fathers have not been socialized to be primary or part-time custodians, and parenting is still considered largely women’s work (Mauer & Pleck, 2006). In general, divorced men are not prepared to assume greater responsibility for parenting even on a part-time basis (Madden-Derdich & Leonard, 2000; Parke & Brott, 1999). For divorced fathers, identities, roles, and functions as a parent become significantly altered and potentially ambiguous, resulting in markedly high stress (Braver, Griffin, Cookston, Sandler, & Williams, 2005). Research shows that divorced fathers report substantially higher levels of parental role strain compared to married fathers (Simon, 1992), and fathers’ role strain is associated with poor psychological health (Umberson & Williams, 1993). We hypothesized that role overload reported by fathers would be associated with poor quality parenting.

Daily hassles and stress

In general, episodic and chronic life stresses experienced by parents predict compromised parenting (Conger, Patterson, & Ge, 1995; Deater-Deckard, 1998). For marital separation in particular, the marked and significant changes in parenting routines, residence, occupational, and social roles during marital separation directly interfere with effective parenting practices (Capaldi & Patterson, 1991; Simons & Associates, 1996; Simons et al., 1993). We hypothesized that potentially stressful changes and events in family, work, and social arenas would be associated with reduced parenting effectiveness.

Social Support and the Stress-Buffering Hypothesis for Divorced Fathers

Social support is particularly salient in the study of divorced fathers because divorce itself can change social networks; further, characteristics of men’s networks, in general, differ from women’s social networks. Divorce significantly reshapes social networks, often leading to some degree of atrophy, lower density, and shorter duration of social ties (Sprecher et al., 2006). Studies also show that men have a higher reliance on friends and extended kin networks for socioemotional support than women (Duran-Aydintug, 1998; Eggebeen, Snyder, & Manning, 1996; Milardo, 1987). That is not to say that women do not rely on friends and kin; rather, on average, women depend less on informal support than men. Because we know little about divorced fathers’ social support, it is important to see who fathers of differing custody status turn to for parenting needs over time.

Finally, we proposed that social support would buffer the negative impact of common stressors for divorced fathers in predicting their quality parenting. The model in Figure 1 is grounded in the ‘‘stress-buffering’’ framework of social support (Schwarzer & Leppin, 1991; Thoits, 1995). This model is based on studies of stress and health that have shown a persons’ appraisal of social support tends to buffer or moderate effects of stress. In the face of stressful conditions, individuals with high levels of social support are buffered from negative effects of stress; however, individuals with low levels of support are more likely to experience negative impacts of stress. Finally, we attempted to advance prior research by focusing on social support for parenting. Fathering outcomes were measured as both coercive and prosocial parenting practices, domains that have been theoretically specified predictors of children’s developmental outcomes (Reid, Patterson, & Snyder, 2002; Simons & Associates, 1996).

Analytic Strategy

Several approaches were used to evaluate the theoretical model. First, we employed a county-representative longitudinal sample of divorced fathers of young children and restricted the sample to fathers reporting contact and parenting interaction with their children. We next conducted a series of custody contrasts on reported sources of social support and potential differences in the levels of specific parenting support domains of emergency and nonemergency child care, practical advice, and financial assistance for parenting. We then tested for differences on parenting outcomes and predictors. Stress-buffering hypotheses were tested in multivariate regression models specifying statistical moderation of social support or interactions among support and family stressors variables predicting parenting.

Method

Two hundred thirty recently divorced fathers participated in the Oregon Divorced Father Study (ODFS). The sample consisted of 31 (14%) full custody, 125 (54%) shared custody, and 74 (32%) no custody fathers. Full custody was defined as sole legal or sole physical custody, shared custody as joint legal or joint physical custody, and no custody as no legal or physical custody rights. Fathers were recruited from a large metropolitan county via public court records. Fathers with children between the ages of 4 and 11 years and a divorce decree date occurring within 24 months from the time of recruitment were eligible for participation. Fathers were asked to participate in the study through letters describing the nature of the project and full explanation of study activities. Fathers’ age ranged from 22.9 to 63.4 years (M = 37.8, SD = 7.7), education ranged from 1 = less than 8th grade to 13 = advanced doctorate (M = 7.2, SD = 2.9), and annual income ranged from 1 = <$5,000 to 10 = >$100,000 (M = 5.4, SD = 2.2). Census data and county police-call data were matched to addresses of participants and nonparticipants. Analysis of neighborhood characteristics including police-call type and frequency, unemployment, homeownership, poverty, and racial makeup indicated nonsignificant differences between participants and nonparticipants. Thirteen percent of the fathers self-identified as racial minorities and 17% identified their children as racial minorities.

One hundred and eighty (78%) of the fathers invited to participate chose to and were able to enroll the focal child. Court records were screened for children within the targeted age range. We randomly selected the focal child if there was more than one eligible child. If the father could not enroll that child, we randomly selected the next eligible child for potential participation until we either enrolled a child or exhausted all eligible children. We found that 92% of full custody and 95% of the shared custody fathers enrolled the targeted focal child, whereas 41% of no custody fathers enrolled the focal child. Therefore, not all full and shared custody fathers had their children participate in the center assessments. However, all fathers, including no custody, filled out questionnaire and interview data on their children’s behavior and their own parenting practices. Therefore, of the original 230 fathers, the sample was restricted to 218 fathers (95%) who reported they had contact with and were able to enroll their child in the study or report on their parenting interactions with their child, or both. Of those children, 46% were girls and 54% were boys. The mean age of the focal child in the present analyses was 7.59 years (SD = 1.96).

Response rates were calculated using methods outlined by Braver and Bay (1992) for court records-based studies in order to weight the sample for selection biases. In total, 867 recruitment letters were sent to fathers, resulting in a located and eligible sampling frame of 572 potential fathers. The overall response rate for participation was 40% for the sample, 55% for full custody fathers, 41% for shared custody fathers, and 35% for no custody fathers. The county population consisted of 10% full custody fathers, 54% shared custody, and 37% no custody fathers. Thus, the ODFS sample overrepresented full custody fathers and slightly underrepresented no custody fathers. In order to obtain data that is county representative of divorced fathers in the sampling frame, we compensated for potential selection biases (Braver & Bay). In short, weighting procedures first corrected for under- or oversampling by comparing the proportions of custody types in the county to proportions of custody types in the sample. Second, a correction for potential selection bias was made on the basis of the participation rates for each custody type. The higher the participation for any custody type, the more representative the sample is for that group; conversely, the lower the participation rate, the higher the threat of selection bias. Therefore, the final weights adjusted for over-or undersampling and for participation rate by custody group. Specific procedures for participation and weighting are provided in the Appendix.

Data were collected during two structured center visits. Each center visit took approximately 2.5 hr to complete. Participants were provided child care, transportation, and a meal if requested, and were paid approximately $25 an hour for their time. Data for the current report were examined across two waves of the longitudinal study: Time 1 and the Time 2 nine-month follow-up assessment. The retention rate for Time 2 was 84%. Attrition analyses revealed no significant differences among fathers retained in the study and those lost to follow up on any of the Time 1 outcomes, key predictors, or control variables, with the exception of the role overload measure. Fathers retained reported higher levels of role overload (M = 2.90, SD = 0.77, and M = 2.49, SD = 0.79, respectively, t =2.13).

Measures

Data were collected with a multiple-method assessment battery. Fathers’ reports of stressors were collected from paper-and-pencil questionnaires, face-to-face interviews, and self-administered computerized questionnaires. Data on parenting behaviors were obtained via questionnaires and observational coding of structured father-child interaction tasks.

Social Support for Parenting

Parenting support was measured with the Parenting Support Index PSI; (DeGarmo & Bryson, 2000), a 24-item questionnaire rated on a 5-point Likert scale from 0 (not at all/not applicable) to 4 (a great deal). Fathers reported the amount of support received within four domains: emergency child care (e.g., if you get sick, have appointments, or have to work overtime), nonemergency child care (e.g., need time to do something fun or relaxing), practical parenting assistance (e.g., advice, doctor referrals, help with doctor appointments, driving to and from daycare/school), and financial assistance with parenting. Each domain was answered for six different relationship types: new partner, relative(s), friend(s), neighbor(s), coworker(s), and former spouse. The total index score was the sum of items in each doman (Cronbach’s α = .82).

Stressors

Conflict

Two indicators from the Barriers to Parental Contact questionnaire (Braver et al., 1993) measured conflict with former spouse. General conflict was an 8-item scale of the father’s reported conflict with former spouse and amount of conflict exposure of the child. Items were rated on a 3-point scale from true to false (e.g., child never sees my ex-wife and me arguing; knows that my ex-wife and I argue or disagree a lot; ex-wife and I are often mean to each other when child is around; often sees arguing; etc., α = .83). Parenting conflict was a report of the amount of conflict associated with coparenting or shared custody relationships specifically computed from five items rated on a 5-point scale from 1 (did not happen) to 5 (happened very often) (e.g., You and your ex-wife argued about moral values related to raising child; discipline practices; activities done with child [e.g., watching TV, selecting movies, wearing bicycle helmets, etc.]; scheduling pickup and drop-off, α = .76).

Role overload

Fathers completed 13 items from the Role Overload Scale (Crouter, Bumpus, Head, & McHale, 2001). Items were rated on a 5-point scale from strongly agree to strongly disagree (e.g., feelings of being overwhelmed by multiple commitments and not having enough time for themselves, there are too many demands on my time, I never seem to get caught up, many times I have to cancel commitments, ODFS α = .93).

Daily hassles

Daily stress measured as family, work, or life hassles were indexed from the Family Events Checklist questionnaire (Patterson, 1982), a 25-item checklist rated on a 4-point scale from 1 (event did not occur) to 4 (event occurred, very negative effect). Sample items included child care problems, stressful work situations, change in financial situation, disagreement with neighbor, and tension between two or more family members not involving you concerning past or present conflict. The items were recoded either 0 or 1 for event occurrence and the final score was the sum of the items indexing cumulative risk (Turner & Wheaton, 1995).

Parenting Quality

Parenting was measured as two specific constructs ‘‘coercive’’ and ‘‘prosocial’’ parenting observed during father-child interaction and reported by the father regardless of whether the child was enrolled in the study. The parenting constructs were assessed from a total of 24 min of videotaped interaction scored across four structured interaction tasks during the father-child visit: a refreshment task (5 min), a problem-solving discussion focusing on a parenting issue (7 min), a play task (7 min), and an academically challenging teaching task (5 min).

For live in vivo interactions, trained observers scored behaviors with the Family and Peer Process Code (FPP; Stubbs, Crosby, Forgatch, & Capaldi, 1998), thereby obtaining information on discrete positive, negative, and neutral behaviors in real time along with information on the initiator, recipient, sequence, content, and affect of behaviors. Fifteen percent of videotapes were randomly selected for blind reliability checks. Cohen’s Kappa, an indicator of coder agreement above chance, was .79 for content (87% agreement) and .80 for affect (95% coder agreement).

Coercive parenting

Three specific indicators comprised the coercive parenting construct. Punitive discipline was a 5-item index rated from 1 (never or almost never) to 5 (always or almost always) in response to the question, ‘‘When (focal child) misbehaves, how often do you . . .?’’ Items included raise your voice/scold, yell, spank on bottom, slap, or hit (α = .85). The second indicator was the rate-per-minute of aversive fathering scored from the FPP coding system using both verbal and physical aggression of the father directed to the child. The third indicator was a 5-item scale of the coder’s impression of harsh discipline rated after scoring FPP discrete behaviors. Rated from 1 (very untrue) to 5 (very true), items included overly strict, authoritarian, expressed hostility during discipline, used nagging, hovered too closely, and used inappropriate discipline (α =.86). To compute a composite construct combining the Likert scales with rate-per-minute counts, each indicator needed to be rescaled to a standard metric. For ease of interpretation, we chose a metric of 0 to 1 because no mean-level information is lost because of standardization as would be the case with combining standardized Z scores. Therefore, Likert-type items ranging from 1 to 5 were recoded 0 to 1 before averaging. Because there are no a priori minimum and maximum values for frequency counts comprising the FPP rate per minute, the coded data were bounded by the minimum and maximum value across time in order to rescale from 0 to 1, resulting in a composite ratio level construct from 0 to 1.

Prosocial parenting

Two Likert-type observational scales were computed from global ratings of prosocial parenting following microsocial scoring of father-child behaviors. Positive involvement was obtained from 14 items rated after each father-child interaction task. Items included ratings on how much the parent treated the child with warmth, empathy, affection, and respect; maintained good eye contact and interactive posture; and so on (α = .94). Skill encouragement was based on ratings of fathers’ ability to promote children’s skill development through contingent encouragement and scaffolding strategies observed during the teaching and construction play tasks. In both tasks, the child is given challenging problems and the father is asked to assist. The scale includes 11 items such as breaks task into manageable steps, reinforces success, prompts, and corrects appropriately (α = .92). Scales were rescored 0 to 1 and averaged.

Control Variables

Several covariates were included that are theoretically relevant to divorce adjustment and parenting. Socioeconomic status (SES) is related to higher levels of effective parenting (Bornstein & Bradley, 2003) and resources are associated with fathers’ custody status (Arditti, 1992). SES was a mean of three standardized scores. Education was measured with years of schooling completed ranging from 1 (<8th grade) to 13 (postgraduate training). Occupation ranged from 1 to 9 using the Hollingshead Four Factor Index of Social Status (Hollingshead, 1975). Income was measured by annual categories ranging from 1 (less than $5,000) to 10 (more than $100,000). Antisocial characteristics and depression are also key covariates of effective parenting and conflict (Patterson, 1982; Patterson & Capaldi, 1991; Simons & Associates, 1996). Antisocial personality was measured with the Acting Out scale of the three-scale Minnesota Multiphasic Personality Inventory (MMPI-TRI; Swanson, Templer, Streiner, Reynolds, & Miller, 1995) consisting of 20 ‘‘yes/no’’ items (e.g., at times feel like picking a fight with someone, I can easily make people afraid of me and sometimes do it for fun, suspended from school one or more times, in trouble with the law, ODFS α = .81). Depressed mood was measured with the Center for Epidemiological Studies Depression Scale (Radloff, 1977), a 20-item symptom-oriented index (e.g., felt depressed, fearful, lonely) rated on a 4-point scale indicating frequency during previous week, ranging from 0 (rarely or none, 0 – 1 day) to 3 (most or all of the time, 5 – 7 days)(α = .88). Additional covariates included: father contact measured as number of days and overnight visits per month, time measured as months since decree, repartnering status measured as cohabiting with an intimate partner for 3 or more months, age, and sex of child.

Results

We first examined responses to the question ‘‘Who do you turn to first when you need help regarding your children?’’ Categories are shown in Table 1 by custody status. The modal categories for full custody fathers were new romantic partner and relatives at Time 1 and Time 2. Shared custody fathers showed similar reliance on new partners and relatives, with roughly a quarter of fathers relying on the former spouse for support at Time 1 and Time 2. No custody fathers were most likely to rely on relatives. We next examined means, standard deviations, and significant differences on the PSI scores (Table 2). Although there were differences exhibited in the sources of support by custody, it was interesting to find that fathers did not differ on the amount of support for parenting reported at baseline or for change over time. This was true for the total index score as well as for each of the subscales.

Table 1.

Fathers’ Report of Primary Source of Support for Parenting Over Time by Custody Status

	Full Custody (%)	Shared Custody (%)	No Custody (%)
Time 1 custody status
Former spouse	0.0	23.8	20.0
Other relative	29.0	31.7	49.2
New partner	48.4	33.7	18.5
Friend or coworker	16.1	8.9	6.2
Other: (professional, God)	6.5	2.0	6.2
Time 2 custody status
Former spouse	6.9	26.1	19.1
Other relative	37.9	29.3	42.6
New partner	31.0	34.8	25.5
Friend or coworker	24.1	9.8	10.6
Other: (professional, God)	0.0	0.0	2.1

Table 2.

Means, Standard Deviations, and Custody Comparisons for PSI Total and Subscales

	Full Custody, M (SD)	Shared Custody, M (SD)	No Custody, M (SD)
T1 PSI				F (2, 217)
T1 emergency child care	6.84 (3.59)	6.79 (3.71)	6.33 (3.69)	0.39
T1 nonemergency child care	5.41 (3.44)	5.48 (3.85)	4.51 (3.18)	1.61
T1 practical (advice, carpool)	5.29 (3.12)	6.10 (3.64)	5.91 (3.17)	0.74
T1 financial assistance	4.17 (3.02)	4.30 (3.60)	4.26 (3.56)	0.02
T1 total index score	21.42 (11.96)	22.06 (13.32)	20.38 (11.50)	0.80
Δ PSI				F (2, 176)
Δ emergency child care	−0.33 (3.03)	0.12 (3.37)	−0.09 (4.37)	0.38
Δ nonemergency child care	0.21 (3.25)	−0.03 (3.40)	1.00 (4.27)	1.21
Δ practical (advice, carpool)	0.44 (2.84)	−0.02 (3.29)	0.45 (3.58)	0.41
Δ financial assistance	0.16 (2.96)	−0.35 (3.20)	0.38 (4.41)	0.70
Δ total index score	1.77 (9.00)	0.34 (12.94)	1.46 (14.94)	0.08

Note. PSI = parenting support index; T1 = Time 1; Δ = Change over 9 months.

We next examined control variables, stress measures, and parenting outcomes (Table 3). As expected, fathers significantly differed on the amount of time spent with children according to custody status. Regarding stress, there was consistent evidence that full custody fathers were more stressed than shared and no custody fathers. Full custody fathers reported greater role strain compared to no custody fathers, higher levels of daily hassles compared to shared and no custody fathers, and higher levels of conflict compared to shared custody fathers. For parenting, shared custody fathers scored higher in prosocial parenting compared to full custody fathers, but no differences were obtained on coercive parenting. For change over time, fathers exhibited no significant differences with the exception of full custody fathers marginally decreasing in daily hassles compared with shared and no custody fathers. No differences in parenting over time were obtained. The current data suggest that divorced fathers differ in time spent with children as a function of custody, but on the average, have similar levels of parenting quality over time.

Table 3.

Means, Standard Deviations, and Custody Comparisons of Control Variables, Father Stressors, and Parenting

	(1) Full Custody, M (SD)	(2) Shared Custody, M (SD)	(3) No Custody, M (SD)	F (2, 216)	Significant Contrasts	R2
T1 father socioeconomic status	−0.10 (0.72)	0.08 (0.70)	−0.00 (0.90)	0.79
T1 proportion repartnered	0.45 (0.51)	0.30 (0.46)	0.24 (0.43)	2.72†
T1 father contact	20.45 (5.99)	14.06 (5.60)	10.02 (5.91)	25.01***	1 > 2 > 3	.22
T1 father acting out	5.89 (4.52)	4.24 (3.03)	5.37 (3.97)	3.43*	1 > 2	.04
T1 father depression	12.34 (7.99)	11.50 (8.68)	14.30 (9.40)	2.03
T1 role overload	3.12 (0.94)	2.89 (0.76)	2.71 (0.69)	3.39*	1 > 2, 3	.03
T1 daily hassles	13.95 (7.14)	11.19 (5.66)	10.63 (4.79)	4.07**	1 > 2, 3	.04
T1 coparental conflict	0.25 (0.18)	0.19 (0.16)	0.24 (0.16)	3.52*	3 > 2	.04
T1 coercive parenting	0.23 (0.12)	0.18 (0.10)	0.19 (0.11)	2.22
T1 prosocial parenting	0.81 (0.11)	0.85 (0.08)	0.83 (0.08)	3.43*	2 > 1	.05
Δ role overload	−0.28 (0.46)	−0.15 (0.60)	−0.00 (0.51)	2.70†
Δ daily hassles	−1.87 (6.05)	−0.33 (6.41)	0.34 (5.46)	1.22
Δ conflict	−0.01 (0.14)	−0.02 (0.11)	−0.01 (0.13)	0.29
Δ coercive parenting	0.01 (0.15)	−0.01 (0.09)	−0.01 (0.10)	0.34
Δ prosocial parenting	−0.02 (0.12)	−0.02 (0.12)	−0.02 (0.08)	0.00

Note. PSI = parenting support index; T1 = Time 1; Δ = Change over 9 months.

^***p < .001

^**p < .001.

^*p < .05.

^†p < .10.

The next set of analyses tested stress-buffering hypotheses using a series of hierarchical regressions for Time 1 and then for change over time. In each of the subsequent models, we entered dummy-coded variables for no custody and full custody effects, and contrasted these with shared custody fathers as the comparison group, because they are identified as the least stressed in the reviewed literature. The first set of models regressed coercive parenting and prosocial parenting on control variables, Time 1 stressors, and Time 1 total parenting support as the first-order predictors. The second block of predictors used stepwise entry of the centered cross-products to test for moderating effects of interaction terms as second-order predictors using regression approaches for testing interactions (Cohen, Cohen, West, & Aiken, 2003). Therefore, all possible buffer effects were tested, but only significant predictors were added to the model. Alpha was set at .10 to report buffer effects at p < .10 since we hypothesized direction of effects. Results and standardized betas for Time 1 are shown in Table 4.

Table 4.

Time 1 Parenting Regressed on Stressors and Hypothesized Moderators

	T1 Coercive Parenting		T1 Prosocial Parenting
	β	Adjusted R2	β	Adjusted R2
Block 1 predictors		.07		.16
Father socioeconomic status	.00		.19*
Father repartnered	−.07		−.24**
Father contact	−.01		.11
Months since decree	−.09		.05
Age of child	.04		−.22**
Sex of child	.04		.07
Father antisocial	.30***		−.08
Father depression	−.03		−.11
No custody	−.07		−.05
Full custody	.17*		−.17†
T1 conflict with ex-wife	.16*		−.14†
T1 role overload	−.05		.01
T1 daily family stress	−.06		.19†
T1 support for parenting	.10		.03
Block 2 significant moderators		.07		.18
T1 Support × T1 Conflict	—		.17*

Note. T1 =Time 1. Betas are standardized coefficients. All possible Stress × Support interactions were tested using stepwise entry. Therefore, only interactions adding significant explained variance are entered in the model.

^***p < .001.

^**p < .001.

^*p < .05.

^†p < .10.

Consistent with coercion theory (Patterson, 1982), father’s acting out was a strong predictor of coercive parenting (β= .30, p < .001) but was not associated with prosocial parenting. Consistent with developmental literature, higher SES was associated with higher levels of prosocial parenting (β= .19, p < .05), and repartnering status was associated with lower levels of prosocial parenting (β= −.24, p < .01), suggesting that repartnering interfered with effective parenting, but repartnering was not associated with higher levels of coercive parenting. Among the stressors, coparental conflict was predictive of higher levels of coercive parenting (β= .17, p < .05) and lower levels of prosocial parenting (β = −.17, p < .10), controlling for antisocial characteristics of the father. Counter to expectations, however, there were no main effects for role overload or daily hassles, with daily hassles predicting prosocial parenting in the opposite direction expected (β= .20, p < .10). Exploratory analyses indicated that this unexpected marginal effect was only present when controlling for amount of father contact with child.

Among stepped-in moderators, the main effect of conflict was not buffered by support in predicting coercive parenting at Time 1 and was significantly buffered by conflict effects on prosocial parenting (β= .17, p < .05). The positive coefficient meant that for fathers with higher levels of social support for parenting, there was a significantly more positive relationship between conflict and prosocial parenting, or conversely stated, there was a less negative impact between conflict and prosocial parenting for fathers with higher parenting support compared with fathers with lower levels of parenting support.

In the final set of models (Table 5), we evaluated the buffering hypothesis specified as change over time. We regressed change scores for parenting on Time 1 and change score predictors and then stepped-in moderator effects of parenting support. One can model change by entering Time 1 and Time 2 scores (also known as the autoregressive method) or by entering Time 1 and a difference score. We chose to model Time 1 and the difference score because beta coefficients represent Time 1 controlling for change and vice versa. Both autoregressive and difference score methods provide beta coefficients for change that are statistically equivalent (see Kessler & Greenberg, 1981). The two methods differ only on the interpretation of Time 1 betas. The advantage of the difference score controlling for Time 1 is that one does not have to algebraically manipulate betas to interpret Time 1 effects controlling for change (Kessler & Greenberg).

Table 5.

Change in Parenting Regressed on Change in Stressors and Parenting Support

	Δ Coercive Parenting		Δ Prosocial Parenting
	β	Adjusted R2	β	Adjusted R2
Block 1 predictors		.23		.33
T1 parenting	−.55***		−.60***
Father socioeconomic status	−.07		.17†
Father repartnered	−.09		.13
Father contact	−.01		.04
Months since decree	.04		.15
Age of child	−.11		−.35***
Sex of child	.08		−.04
Father antisocial	−.01		−.10
Father depression	.11		.03
T2 no custody	.05		.03
T2 full custody	.12		−.05
Δ conflict	.06		.02
Δ role overload	.20*		−.09
Δ daily stress	−.08		−.07
Δ parenting support	.06		.02
Block 2 significant moderators		.26		.36
T1 Support × Δ Role Overload	−.12†		.14†
Δ Support × Δ Conflict	−.21*		—
T1 Support × Δ Daily Stress	—		.13†
Δ Support × Δ Daily Stress	—		.20*

Note. T1 = Time 1; T2 = Time 2; Δ = Change over 9 months. Models control for T1 predictors. Betas are standardized coefficients. All possible Stress × Support interactions were tested using stepwise entry.

^***p < .001.

^**p < .001.

^*p < .05.

^†p < .10.

For stress, 9-month increases in role overload were significantly associated with increases in coercive parenting (β = .20, p < .05) but not with reductions in prosocial parenting. Unlike Time 1 models, change in conflict was not an associated change in prosocial parenting. A buffering model became more salient over time as exhibited by moderators entering the change score regressions. Higher levels of Time 1 support for parenting marginally buffered the effects of increased role overload on coercive parenting (β = −.21, p < .06) and increases in parenting support buffered effects of increases in conflict on change in coercive parenting (β = −.21, p < .05). For change in prosocial parenting, buffering effects were found for change in parenting support moderating effects of increases in daily hassles (β = .20, p < .05), and marginal effects for Time 1 parenting support on increases in role overload (β = .14, p < .10) and increases in daily hassles (β = .14, p < .10). Therefore, for fathers with higher levels of social support, the slope of change in conflict and change in hassles were more positive compared to fathers with lower levels of parenting support over time whose stress slopes were more negative or detrimental. The displayed increments in R² for the buffering effects were significant.

Discussion

We know more about effects of fathers’ involvement on the developmental outcomes for children than we know about the factors promoting quality parenting for fathers. We know even less about fathering determinants following divorce, although there has been a recent increased interest in studying divorced and nonresidential fathers. One area that has been understudied is the domain of social support for parenting behaviors and parenting needs of divorced fathers. The present study attempted to address the question of where fathers turn to for support and how this support might mitigate effects of stress accompanying the changes in marital separation.

Unfortunately, fathers tend to disengage from parenting responsibilities for a variety of reasons following divorce. Many of these reasons are associated with a subjective cost-benefit analysis of parenting stresses and continued coparenting conflict (Braver et al., 1993). Given that fathers are vulnerable to markedly high levels of distress initially following the divorce, and given that fathers are generally not prepared to be primary custodians, it is important to understand how support may mitigate disengaged parenting over time because quality divorced father involvement is associated with greater child support compliance (Arditti & Keith, 1995) and better child adjustment (King & Sobolewski, 2006). Additionally, parents eventually tend to reduce their levels of coparental conflict (Braver, Griffin, Cookston, Sandler, et al., 2005).

Descriptively, we first found that no custody fathers had a higher reliance on relatives, full custody fathers had a lower reliance on former spouses, and consistent with prior studies, custodial fathers relied more on new partners compared to noncustodial fathers. The finding that a new partner is important in helping a father adjust to divorce has been well documented, but much less is known about other sources of nonpartner support (Stone, 2002). It will be important in future analyses to examine the role of support relationship types and independent effects of nonmarital/nonromantic sources of support that may be clinically relevant for both recently repartnered fathers and fathers remaining single following separation.

Surprisingly, although we found differences in where fathers turned for advice and practical assistance with parenting, we found no differences among the amounts and reported levels of social support in the specific parenting domains of child care needs. Further, fathers did not differ on change in validated measures of parenting practices in the coercive or prosocial parenting domains as a function of custody status. At Time 1, shared custody fathers exhibited higher prosocial parenting than full custody fathers. This suggests that fathers who eventually divorce are more similar in their parenting behaviors over time, regardless of custody status. It is likely that individual differences may account for variation in parenting quality more than between-group variables such as custody status. Similarly, differences in personality characteristics are known predictors of ability to garner social support and maintain support networks (Schwarzer & Leppin, 1991). Accordingly, although we did not include sociability, we controlled for antisocial personality.

In tests of the stress-buffering model, we found evidence supporting buffering effects of social support; in adition, these effects become more important over time. Increases in role overload and increases in parenting conflict with the former spouse were associated with changes in coercive parenting for those fathers with relatively lower levels of social support. Changes in prosocial parenting, on the other hand, were predicted by a conditional relationship among parenting support and by the experience of daily hassles.

It is important to note that coercive and prosocial parenting domains were distinct. Both are key mechanisms shaping the developmental trajectories of children’s conduct problems following divorce. However, the present data suggest that more visceral or aversive stressors may be linked to coercive parenting, whereas less insulting stressors, daily hassles, are linked with prosocial parenting under conditions of low parenting support. We can only speculate as to why father contact would suppress the effect of daily hassles on prosocial parenting. Perhaps, more engaged and involved fathers may experience greater levels of daily stress. This finding was not found over time and better longitudinal specification is needed to explore this unexpected result.

Implications for Practice

The current findings also suggest that parent-training interventions or clinical practice with fathers focusing on parenting needs would be better informed by paying attention to the fathers’ social contexts. A particular focus should be on interparental conflict over time and father’s experiences of stress and role overload. It is not clear why the beneficial effects of social support as a protective factor were more predictive of quality parenting over time. However, this suggests that clinical intervention may speed up the process and also may be needed early in the divorce process to have optimal impact on buffering fathers’ parenting relationships.

Regarding the ability to garner social support, one might expect to find greater efficacy in parent training by incorporating aspects such as help seeking, problem solving, and interpersonal skill building related to parenting needs. Effects might also be greater for fathers who are relatively more insular compared to fathers with established functional social networks. Experimental trials comparing a social support enhancement component could address this.

Related, effective intervention strategies would encompass aspects that are conditioned by social support such as managing stress and conflict. Unfortuantely, very few such tailored programs exist. Dads for Life (DFL; Braver, Griffin, & Cookston, 2005) is one exception of an evidence-based program designed for postdivorce quality father involvement. DFL includes dimensions focused on commitment to the parenting role, motivation and skills for managing conflict with the former souse, and skills for parenting. Support groups as well as individual modes of intervention can be particularly salient for fathers. Parke and Brott (1999) reported that men who participate in support groups can experience a powerful sense of centeredness, arising from a growing sense of affirmed identity within a community. However, men who join support groups tend to stay active only until their particular problems and concerns are ameliorated and then tend to move on.

One central clinical recommendation is to focus on fathers’ view of the parenting role itself. Although historically fathers are becoming more involved in their children’s lives, fathers still identify with ‘‘breadwinning’’ (Mauer, Pleck, & Rane, 2001), and ‘‘caregiving’’ is still primarily defined as woman’s work. This means that many fathers need to cognitively redefine tasks that are nontraditional for men as still somehow being masculine to reduce threats to their own masculinity (Doucet, 2004; Mauer & Pleck, 2006). Because men continue to be socialized as helpers to mothers, Parke and Brott (1999) recommended that parenting education should begin early in schools to reduce gender stereotypes. Men need to think of being partners not only as helpers to mothers; couples err by neglecting to give parenting the same weight as other domestic chores.

Limitations and Advantages

Although the present sample was county representative, it was a small regional sample. We need to understand social interactional determinants of effective fathering in more diverse samples. It is likely that the factors associated with fathering and postdivorce contexts are culturally specific (Coley, 2001). Presently, we found some marginal effects, and it is also possible that some of the moderating tests were underpowered. Larger samples may better inform specific delineation of stress processes. Given these limitations, however, the present study had advantages of using mixed methods to assess specific parenting practices. The mixed-method approach limits potential response bias in the models. We also employed longitudinal data showing that buffering effects were more apparent over time. It will be important to conduct longer term follow-up evaluations.

Table A1.

Recruitment Sampling Frame (N = 867 County Records)

		No Custody	Shared Custody	Full Custody	Total
Letters sent		363	424	80	867
Not located		−17	−12	−3	−32
Located		346	412	77	835
Ineligible		−134	−108	−21	−263
Located and eligible		212	304	56	572
Refused		−138	−179	−25	−342
Participated/assessed (n)		74	125	31	230
For population weights
Letters sent	867
Not located	32
Refused	342
Participated/assessed	230
Eligible = Letters sent − ineligible	604 = 867 − 263
Corrected = Letters sent − (Ineligible × Nonlocate Rate)	590 = 867 − ((263 + (.05 × 263))
Nonlocation rate (eligible)	.053 = 32/604

Table A2.

Located and Eligible Sampling Frame (n = 572)

	No Custody, n (%)	Shared Custody, n (%)	Full Custody, n (%)
Population	212 (37)	304 (54)	56 (9.8)
ODFS Sample	74 (32)	125 (53)	31 (13.5)
Uncorrected class sample weights	1.15 (Undersampled)	1.01	0.73 (Oversampled)
		χ2 (2) = 7.89, p < .05

Table A3.

Sample Proportions, Estimated Population Proportions, and Sample Weights

		ODFS Sample Proportions
	Estimated
	Population
Conservative	Proportion	No Custody	Shared Custody	Full Custody	Total
Participated (A)	.381 = (230/604)	.322 = (74/230)	.543 = (125/230)	.135 = (31/230)	1.00
Not located	.053 = (32/604)	.531 = (17/32)	.375 = (12/32)	.093 = (3/32)	1.00
Refused	.566 = (342/604)	.403 = (138/342)	.523 = (179/342)	.073 = (25/342)	1.00
Liberal (corrected for ineligible)
Participated	.390 = (230/590)
Not located	.050 = (30/590)
Refused	.579 = (342/590)
(B) Weighted average sample         proportion (conservative)		(.381 × .322)   +(.053 × .531)   +(.566 × .403)   = .379	(.381 × .543)   +(.053 × .375)   +(.566 × .523)   = .523	(.381 × .135)   +(.053 × .093)   +(.566 × .073)   = .098	1.00
(C) Weighted average sample        proportion (liberal)		(.390 × .322)   +(.050 × .531)   +(.579 × .403)   = .385	(.390 × .543)   +(.050 × .375)   +(.579 × .523)   = .533	(.390 × .135)   +(.050 × .093)   +(.579 × .073)   = .099	1.00
Conservative weights (B/A)		1.177 = .379/.322	.963 = .523/.543	.725 = .098/.135
Liberal weights (C/A)		1.195 = .385/.322	.981 = .533/.543	.733 = .099/.135
Average weights		1.186	.972	.729
N corrected = (n/Σ weights)	= (230/231.41) = .993	1.177	.965	.723

Note. Conservative Participation Rate is (Participated/Located − Ineligible). Liberal Participation Rate is (Participated/(Located − Ineligible) – (Ineligible Rate × Nonlocation Rate)).